Zheng Niu

Feature Selection of Time Series MODIS Data for Early Crop Classification Using Random Forest: A Case Study in Kansas, USA

Currently, accurate information on crop area coverage is vital for food security and industry, and there is strong demand for timely crop mapping. In this study, we used MODIS time series data to investigate the effect of the time series length on crop mapping. Eight time series with different lengths (ranging from one month to eight months) were tested. For each time series, we first used the Random Forest (RF) algorithm to calculate the importance score for all features (including multi-spectral data, Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and phenological metrics). Subsequently, an extension of the Jeffries–Matusita (JM) distance was used to measure class separability for each time series. Finally, the RF algorithm was used to classify crop types, and the classification accuracy and certainty were used to analyze the influence of the time series length and the number of features on classification performance; the features were added one by one based on their importance scores. Results indicated that when the time series was longer than five months, the top ten features remained stable. These features were mainly in July and August. In addition, the NDVI features contributed the majority of the most significant features for crop mapping. The NDWI and data from multi-spectral bands also contributed to improving crop mapping. On the other hand, separability, classification accuracy, and certainty increased with the number of features used and the time series length, although these values quickly reached saturation. Five months was the optimal time series length, as longer time series provided no further improvement in the classification performance. This result shows that relatively short time series have the potential to identify crops accurately, which allows for early crop mapping over large areas.

The Potential of Time Series Merged from Landsat-5 TM and HJ-1 CCD for Crop Classification: A Case Study for Bole and Manas Counties in Xinjiang, China

Time series data capture crop growth dynamics and are some of the most effective data sources for crop mapping. However, a drawback of precise crop classification at medium resolution (30 m) using multi-temporal data is that some images at crucial time periods are absent from a single sensor. In this research, a medium-resolution, 15-day time series was obtained by merging Landsat-5 TM and HJ-1 CCD data (with similar radiometric performances in multi-spectral bands). Subsequently, optimal temporal windows for accurate crop mapping were evaluated using an extension of the Jeffries–Matusita (JM) distance from the merged time series. A support vector machine (SVM) was then used to compare the classification accuracy of the optimal temporal windows and the entire time series. In addition, different training sample sizes (10% to 90% of the entire training sample in 10% increments; five repetitions for each sample size) were used to investigate the stability of optimal temporal windows. The results showed that time series in optimal temporal windows can achieve high classification accuracies. The optimal temporal windows were robust when the training sample size was sufficiently large. However, they were not stable when the sample size was too small (i.e., less than 300) and may shift in different agro-ecosystems, because of different classes. In addition, merged time series had higher temporal resolution and were more likely to comprise the optimal temporal periods than time series from single-sensor data. Therefore, the use of merged time series increased the possibility of precise crop classification.

Geostatistical modeling using LiDAR-derived prior knowledge with SPOT-6 data to estimate temperate forest canopy cover and above-ground biomass via stratified random sampling

Forest canopy cover (CC) and above-ground biomass (AGB) are important ecological indicators for forest monitoring and geoscience applications. This study aimed to estimate temperate forest CC and AGB by integrating airborne LiDAR data with wall-to-wall space-borne SPOT-6 data through geostatistical modeling. Our study involved the following approach: (1) reference maps of CC and AGB were derived from wall-to-wall LiDAR data and calibrated by field measurements; (2) twelve discrete LiDAR flights were simulated by assuming that LiDAR data were only available beneath these flights; (3) training/testing samples of CC and AGB were extracted from the reference maps inside and outside the simulated flights using stratified random sampling; (4) The simple linear regression, ordinary kriging and regression kriging model were used to extend the sparsely sampled CC/AGB data to the entire study area by incorporating a selection of SPOT-6 variables, including vegetation indices and texture variables. The regression kriging model was superior at estimating and mapping the spatial distribution of CC and AGB, as it featured the lowest mean absolute error (MAE; 11.295% and 18.929 t/ha for CC and AGB, respectively) and root mean squared error (RMSE; 17.361% and 21.351 t/ha for CC and AGB, respectively). The predicted and reference values of both CC and AGB were highly correlated for the entire study area based on the estimation histograms and error maps. Finally, we concluded that the regression kriging model was superior and more effective at estimating LiDAR-derived CC and AGB values using the spatially-reduced samples and the SPOT-6 variables. The presented modeling workflow will greatly facilitate future forest growth monitoring and carbon stock assessments for large areas of temperate forest in northeast China. It also provides guidance on how to take full advantage of future sparsely collected LiDAR data in cases where wall-to-wall LiDAR coverage is not available from the perspective of geostatistics.

Reconstruction of Daily 30 m Data from HJ CCD, GF-1 WFV, Landsat, and MODIS Data for Crop Monitoring

With the recent launch of new satellites and the developments of spatiotemporal data fusion methods, we are entering an era of high spatiotemporal resolution remote-sensing analysis. This study proposed a method to reconstruct daily 30 m remote-sensing data for monitoring crop types and phenology in two study areas located in Xinjiang Province, China. First, the Spatial and Temporal Data Fusion Approach (STDFA) was used to reconstruct the time series high spatiotemporal resolution data from the Huanjing satellite charge coupled device (HJ CCD), Gaofen satellite no. 1 wide field-of-view camera (GF-1 WFV), Landsat, and Moderate Resolution Imaging Spectroradiometer (MODIS) data. Then, the reconstructed time series were applied to extract crop phenology using a Hybrid Piecewise Logistic Model (HPLM). In addition, the onset date of greenness increase (OGI) and greenness decrease (OGD) were also calculated using the simulated phenology. Finally, crop types were mapped using the phenology information. The results show that the reconstructed high spatiotemporal data had a high quality with a proportion of good observations (PGQ) higher than 0.95 and the HPLM approach can simulate time series Normalized Different Vegetation Index (NDVI) very well with R2 ranging from 0.635 to 0.952 in Luntai and 0.719 to 0.991 in Bole, respectively. The reconstructed high spatiotemporal data were able to extract crop phenology in single crop fields, which provided a very detailed pattern relative to that from time series MODIS data. Moreover, the crop types can be classified using the reconstructed time series high spatiotemporal data with overall accuracy equal to 0.91 in Luntai and 0.95 in Bole, which is 0.028 and 0.046 higher than those obtained by using multi-temporal Landsat NDVI data.

Characterizing Radiometric Attributes of Point Cloud Using a Normalized Reflective Factor Derived From Small Footprint LiDAR Waveform

This paper presents a reflectance-like coefficient, normalized reflective factor (NRF) to characterize the radiometric attributes of point cloud generated from small footprint light detection and ranging (LiDAR) waveform data. The NRF is defined as a normalized ratio between the energy of emitted laser beam and the peak in return waveform in conjunction with the atmospheric attenuation and observation geometry. Based on the Gaussian parameters of the emitted and return waveforms, NRF is calculated with an empirical atmospheric model and user-defined standard observation geometry. To correct the radiometric measurement of point cloud in multipeak waveform, a semi-physical-based method is adopted to enhance the NRF of point cloud generated from multipeak waveform. Experiments are conducted with small footprint LiDAR waveform data acquired by RIEGL LMS-Q560. A curve-fitting-based approach is applied to decompose LiDAR waveform into three-dimensional (3-D) coordinates of point cloud, and the NRF are calculated using the Gaussian parameters of both emitted and return waveforms. The visualization of the radiometric attributes of point cloud data is carried out over the overlapping areas between different flight strips, it suggests that the NRF over overlapping area is much smooth than the normalized intensity. Quantitative comparison with Hyperion data indicates that the NRF has much higher correlation with surface reflectance than the normalized intensity data. Standard deviations of NRF and the normalized intensity of different land cover patches are analyzed to assess the homogeneity of the radiometric data. It is observed that NRF has less variability than the normalized intensity within the same land cover patches. Point cloud of two sample trees is also selected to assess the performance of the “sub-footprint” effect correction. It is observed that the proposed approach reduced the variability of radiometric attributes over tree canopies with increasing NRF values; which means the “sub-footprint” effect is mitigated. In summary, the proposed NRF can serve as a promising indicator to characterize radiometric attribute of LiDAR point cloud.

Combined Use of Airborne LiDAR and Satellite GF-1 Data to Estimate Leaf Area Index, Height, and Aboveground Biomass of Maize During Peak Growing Season

A fast and efficient estimation of crop biophysical parameters is significantly important in many agricultural, ecological, and meteorological applications. This study investigated the potential of airborne LiDAR and satellite GF-1 data for estimating three biophysical parameters of maize: 1) leaf area index (LAI); 2) average canopy height (Hcanopy); and 3) aboveground biomass (AGB) during the peak growing season. First, classification data of maize was produced using normalized surface height, GF-1 NDVI, and terrain slope through decision-making. Second, four representative remotely sensed (RS) metrics which have been widely used in forest studies were tested to develop multiplicative models with similar shapes for estimating each biophysical parameter of maize, respectively. Third, the estimation results were obtained and validated through leave-one-out cross-validation method yielding a root-mean-square error (rmse) of 0.37 for LAI, 0.17 m for Hcanopy, and 0.49 kg/m2 for AGB. Finally, contributions to the estimation models from each RS metric were analyzed, and spatial patterns of the biophysical parameters across the entire study area were mapped. Based on these results, the following conclusions were drawn. 1) The four selected metrics from airborne LiDAR and satellite GF-1 data are also applicable and promising in estimating biophysical parameters of maize during the peak growing season. 2) Multiplicative model was proved to be a fast, simple but effective alternative by combining LiDAR-derived structure information and spectral content from GF-1 NDVI. These conclusions provide valuable information for estimation of biophysical parameters of maize during the peak growing season.

Evaluation of Orthomosics and Digital Surface Models Derived from Aerial Imagery for Crop Type Mapping

Orthomosics and digital surface models (DSM) derived from aerial imagery, acquired by consumer-grade cameras, have the potential for crop type mapping. In this study, a novel method was proposed for extracting the crop height from DSM and for evaluating the orthomosics and crop height for the identification of crop types (mainly corn, cotton, and sorghum). The crop height was extracted by subtracting the DSM derived during the crop growing season from that derived after the crops were harvested. Then, the crops were identified from four-band aerial imagery (blue, green, red, and near-infrared) and the crop height, using an object-based classification method and a maximum likelihood method. The results showed that the extracted crop height had a very high linear correlation with the field measured crop height, with an R-squared value of 0.98. For the object-based method, crops could be identified from the four-band airborne imagery and crop height, with an overall accuracy of 97.50% and a kappa coefficient of 0.95, which were 2.52% and 0.04 higher than those without crop height, respectively. When considering the maximum likelihood, crops could be mapped from the four-band airborne imagery and crop height with an overall accuracy of 78.52% and a kappa coefficient of 0.67, which were 2.63% and 0.04 higher than those without crop height, respectively.

Dynamic Monitoring of the Largest Freshwater Lake in China Using a New Water Index Derived from High Spatiotemporal Resolution Sentinel-1A Data

Poyang Lake is the largest freshwater lake in China and is well known for its ecological function and economic importance. However, due to the influence of clouds, it is difficult to dynamically monitor the changes in water surface areas using optical remote sensing. To address this problem, we propose a novel method to monitor these changes using Sentinel-1A data. First, the Sentinel-1A water index (SWI) was built using a linear model and a stepwise multiple regression analysis method with Sentinel-1A and Landsat-8 imagery acquired on the same day. Second, water surface areas of Poyang Lake from 24 May 2015 to 14 November 2016 were extracted by the threshold method utilizing time-series SWI data with an interval of 12 days. The results showed that the SWI threshold classification method could be applied to different regions during different periods with high quantity accuracy (approximately 99%). The water surface areas ranged between 1726.73 km2 and 3729.19 km2 during the study periods, indicating an extreme variability in the short term. The maximum and average values of the changed areas were 875.57 km2 (with a change rate of 35%) and 197.58 km2 (with a change rate of 8.2%), respectively, after 12 days. The changes in the mid-western region of Poyang Lake were more dramatic. These results provide baseline data for high-frequency monitoring of the ecological environment and wetland management in Poyang Lake.

Analyzing the Sensitivity of Crops Classification Accuracy Based on MODIS EVI Time Series and History Ground Reference Data

Abstract An improved spectral profile–based classification method was developed to discriminate corn, alfalfa, and winter wheat in the U.S. state of Kansas. Unlike other classification procedures, this method uses historical field reference data as training samples. An artificial immune network (AIN) algorithm, namely the artificial antibody network (ABNet), was tested as a classifier, combining historical field reference data and moderate-resolution imaging spectroradiometer (MODIS)-enhanced vegetation index (EVI) images. Historical field reference data from the years 2009 to 2012 were used to classify the three crops for 2013 data. A new method was developed to select the purest pixels from cropland data layer (CDL). Historical reference data were used in two different methods to classify crops in 2013: (i) single-year historical data and (ii) multiyear data used in four different combinations. Using method (i), classification was most accurate when the most recent year of training data was utilized. The accuracy of method (ii) increased with the number of years of data used for training the classifier. Results ranged from 81% to 92% overall accuracies, with the exception of the year 2012, where a severe drought created anomalous spectral profiles for all crops in the study area.

Identifying crown areas in an undulating area planted with eucalyptus using unmanned aerial vehicle near-infrared imagery

ABSTRACT In this letter, we propose an identification method of tree crown areas for imagery captured by a near-infrared camera on board an unmanned aerial vehicle platform over an undulating Eucalyptus planting area in Guangdong Province, China. The method extracts crown areas by applying mathematical morphology, unsupervised segmentation based on J-value segmentation, local spatial statistics, and Iterative Self-Organizing Data Analysis Technique Algorithm. Two morphology filters and four segmentation scales were compared between densely and sparsely planted plots as well as sunlit and shaded plots. The opening operation by the window size of 9×9 pixel and segmentation by the seed area sized 65×65 pixel achieved the best performance with overall accuracy of 91%, 93%, 89% and 91% in densely sunlit, sparsely sunlit, densely shaded and sparsely shaded plots.